Re: [RFA] Emit a warning for ineffective set VAR = EXP command

[Tristan Gingold <gingold@adacore.com>]

On May 7, 2012, at 9:38 PM, Joel Brobecker wrote:

>>> This warns about "set variable $j++" presumably -- should the warning be 
>>> disabled for pre/post increments/decrements?
>> 
>> I am not opposed to disable warnings for pre/post inc/dec.
>> But this usage is dubious (the help explicitly mentions VAR=EXP !)
>> 
>> Opinion ?
> 
> I think we should avoid the warning for pre/post inc/dec. This
> type of expression might be a little outside the method proposed
> in our documentation, but I think it's still a perfectly valid
> expression that results in an assignment being performed.

I don't know who should approve this adjustment, but here is the version that deals with pre/post inc/dec.
Note that it still warns for expressions such as i++ * 2.

Tested on set-nowarns.exp.

Tristan.

2012-05-09  Tristan Gingold  <gingold@adacore.com>

	* printcmd.c (set_command): Add pre/post inc/dec.

diff --git a/gdb/printcmd.c b/gdb/printcmd.c
index 79e38f2..fa76296 100644
--- a/gdb/printcmd.c
+++ b/gdb/printcmd.c
@@ -1080,11 +1080,21 @@ set_command (char *exp, int from_tty)
   struct cleanup *old_chain =
     make_cleanup (free_current_contents, &expr);
 
-  if (expr->nelts >= 1
-      && expr->elts[0].opcode != BINOP_ASSIGN
-      && expr->elts[0].opcode != BINOP_ASSIGN_MODIFY
-      && expr->elts[0].opcode != BINOP_COMMA)
-    warning (_("Expression is not an assignment (and might have no effect)"));
+  if (expr->nelts >= 1)
+    switch (expr->elts[0].opcode)
+      {
+      case UNOP_PREINCREMENT:
+      case UNOP_POSTINCREMENT:
+      case UNOP_PREDECREMENT:
+      case UNOP_POSTDECREMENT:
+      case BINOP_ASSIGN:
+      case BINOP_ASSIGN_MODIFY:
+      case BINOP_COMMA:
+	break;
+      default:
+	warning
+	  (_("Expression is not an assignment (and might have no effect)"));
+      }
 
   evaluate_expression (expr);
   do_cleanups (old_chain);